Driverless rides in Paris could begin as an experiment from 2028 under a plan announced by Bolt, carmaker Stellantis and the Métropole du Grand Paris on 23 October 2026. It is a proposal for a carefully assessed service, not the arrival of a robotaxi that Paris residents can order today. The difference matters because some of the biggest obstacles are not software problems at all.
The partners want to test autonomous vehicles initially in up to five municipalities, with an ambition to extend operations more widely by 2030 if the evidence, approvals and economics justify it. They have applied to a European urban-mobility initiative, and the routes and permissions still have to be settled. For riders, the real question is whether the cars can solve awkward journeys that buses and trains do not conveniently cover.
How driverless rides in Paris might work
The planned pilot is aimed at Greater Paris, not simply a sightseeing demonstration on famous central streets. The official announcement points to connections between homes and public transport, employment centres, off-peak journeys and large events. These are familiar frustrations: a train can be fast, yet the trip to the station or the final stretch home can still be inconvenient.
The Métropole du Grand Paris represents Paris and its surrounding municipalities, while Bolt would plan how the autonomous service fits into a transport app people already use. Stellantis would provide vehicles prepared for advanced automated driving. The partners have identified Pony.ai as a possible supplier of the driving system. Calling that a final supplier appointment would be premature; it remains a potential role within the announced plan.
The proposal also raises a practical point frequently missed in glossy driverless-car demonstrations. A vehicle that negotiates a planned test route is different from a service that works reliably throughout a complicated city, in rain, near roadworks, alongside cyclists and around temporary restrictions. A pilot can examine those everyday conditions, but its existence does not prove an eventual commercial operation will be viable.
Why 2028, rather than next month?
Autonomous driving requires more than a computer detecting pedestrians and road signs. Operators need suitable vehicles, approved operating areas, safety arrangements, remote assistance policies and a process for investigating incidents. Local authorities must understand how the cars will interact with buses, taxis, emergency services and people who cannot easily use a smartphone.
The October announcement makes the sequence explicit: choose participating municipalities, agree suitable routes, obtain necessary permissions and then evaluate the operation independently. The partners want evidence about safety, reliability, public acceptance and the wider effects on mobility. The goal of potential expansion by 2030 is an ambition, not a promised schedule or a guarantee that every resident will be able to hail a car.
Bolt has also described a wider programme of autonomous driving partnerships in Europe. Its press-releases page lists a separate announcement about choosing NVIDIA technology for future deployment, as well as an agreement involving Lucid. Those plans are relevant commercial context, but one vehicle programme should not be conflated with the specific Greater Paris pilot. The Paris announcement centres on Stellantis vehicles, city oversight and a potentially different driver technology partner.
The awkward question of who controls the car
A fully driverless service changes the role traditionally performed by a professional driver. Someone still has to decide where a car may operate, when conditions are unsafe, what happens if it stops in an unsuitable place and how passengers obtain help. A transport platform must also explain how a rider would report a problem and how responsibility is divided between the vehicle maker, the system supplier and the service operator.
The official Paris proposal says the trial will be subject to safety, regulatory and data-protection requirements. It does not establish the final answer to every operational question. Those details will be a crucial part of the independent evaluation. Riders will reasonably care about accessibility, luggage, child seats, passenger support, emergency stops and price as much as they care about the underlying AI.
LiveAIWire previously explored research on how self-driving cars explain their predictions and AI announcements about bus delay communications. Making the car more understandable may help engineers and oversight teams examine decisions, but good explanations are not a substitute for dependable road behaviour. Nor can an attractive passenger interface establish safety on its own.
A small service may prove more useful than a big promise
There is a difference between replacing buses and filling gaps around them. The project partners explicitly frame the trial as a complement to existing networks. That could mean a journey to a station or a trip when a conventional service is infrequent, rather than a robotaxi competing with every bus on the same street. Whether such a model reduces traffic, improves access or simply adds vehicles will have to be measured rather than presumed.
The early pilot would generate operating data to inform decisions by Greater Paris. Independent evaluation is particularly important because companies selling the technology also have an interest in wider deployment. Information about service quality, safety incidents, public use and accessibility will help determine whether expansion is warranted.
The next meaningful milestones are not dramatic product videos. They are the selection of actual test areas, a clear timetable, the permissions secured and evidence from running a service with real passengers. Until then, the headline development is a partnership and a concrete proposal to begin trials from 2028, not a public driverless taxi launch.
About the Author
Stuart Kerr is Technology Correspondent at LiveAIWire, covering artificial intelligence, cybersecurity and the social impact of emerging technology. LiveAIWire is an independent, human-led technology publication using AI-assisted research, editorial production and original AI-assisted editorial illustrations under his direction.